應用資料探勘技術提昇CMOS產業需求規劃系統之資料品質研究

Abstract

CMOS Image Sensor產業伴隨智慧行動裝置的成長而持續蓬勃發展，其未來展望透過各項智慧嵌入式系統的創新應用，可望持續帶動相關影像感測元件市場持續成長，由於V公司專注於影像感測元件之後段製程生產與服務並在近年來隨著新事業群向外發展趨勢日益增加，因此對商業智慧平台的整合資訊系統與資料倉儲服務需求也是持續增加，其中以商業智慧平台為主的相關系統的資料正確性更是影響高階主管決策的重要因素，因此「資料品質」已成為當前影響商業智慧資訊系統導入成敗是一項重要課題。 V公司商業智慧平台的需求規劃系統資料取得以多個資料來源為主，將不同來源資料以主題方式分類後，並將資料轉換到資料倉儲內，作為商業智慧平台資料來源的依據。 此外將資料轉換到資料倉儲要要進行資料轉換與過濾方面，運用資料探勘技術是否可以有效提昇需求規劃系統的資料品質問題值得探討與研究，其研究過程為了找出系統資料不正確、錯誤及不一致並產生高品質的資料是非常重要的課題，因此本研究提出以資料探勘的關聯法則與決策樹概念為基礎的自動化建立資料檢驗規則，此新方法比傳統人工建立資料檢規則方法會來的更精確及更有效率並且提昇商業智慧系統的資料品質。

With the growth of smart mobile devices, the CMOS Image Sensor industry continues to flourish, and its future prospects through innovative applications of the wisdom embedded systems, continue to drive the growth of image sensor market. Due to V Company focusing on the image sensor back-end process production and services in recent years, with the new business group’s outside development trend of increasing integration of information systems and data warehousing services, there is a need to enhance the business intelligence platform; a business intelligence platform-based system of data accuracy is an important factor affecting senior management’s decision-making. The impact of "data Quality" on the success or failure of current business intelligence platforms has become an important issue. V Company gets the majority of its data from multiple data sources in their demand planning system which stores different data in the data warehousing, providing a single data source for their business intelligence system. In addition to data conversion and filtering processing to the data warehouse, the use of data mining techniques can effectively improve the data quality of the demand planning system; this approach is worthy of discussion and research. In the research process, determining if the system information is incorrect, finding errors and inconsistencies and producing high-quality data is challenging; this study proposes data mining association rules and decision tree concepts to create inspection rules-based automation. The new test method is more accurate and more efficient than the traditional artificial data for rules method and enhances the business intelligence system data quality.